Safety-Critical Control under Physical Input Limits

This direction is split into other directions: actuator saturation constraints. Its last version stays readable and citable here.

Version 1 of 147 papers59 conceptsfrozen 2026-10-04

In the ideal world of control theory, a robot can always execute the commands its controller issues. If a self-driving car is heading toward a wall, the math assumes the brakes can provide whatever force is necessary to stop in time. In reality, actuators—the motors, thrusters, and brakes that move a system—have hard physical limits. A motor can only provide so much torque; a brake can only apply so much friction.

When these physical limits conflict with safety requirements, traditional safety filters fail. This research direction focuses on bridging the gap between theoretical safety guarantees and the "input-constrained" reality of hardware.

The Safety Filter and the Feasibility Gap

The standard tool for ensuring safety in modern robotics is the Control Barrier Function (CBF). Introduced in foundational work by Aaron D. Ames, Xiangru Xu, and colleagues in Control Barrier Function Based Quadratic Programs with Application to Automotive Safety Systems, a CBF acts as a mathematical "guardrail." It defines a safe set of states and ensures that the system remains "forward invariant"—meaning if it starts safe, it stays safe.

This is typically implemented as a Quadratic Program (QP). The QP takes a desired performance command (like "drive at 60 mph") and filters it into a safe command. However, as Mitchell Black notes in Consolidated Control Barrier Functions: Synthesis and Online Verification via Adaptation under Input Constraints, these frameworks often assume "unlimited control authority."

If the safety filter determines that the only way to stay safe is to decelerate at 20 m/s², but the vehicle's tires can only handle 10 m/s², the QP becomes infeasible. The controller "paints itself into a corner" where no valid physical command can satisfy the safety constraint.

Designing for Feasibility

To prevent this, researchers have moved toward "feasibility-aware" safety. The goal is to ensure that the safe set is not just a region where the robot is safe, but a region from which the robot can remain safe using only the power it actually has.

High-Order Constraints

Wei Xiao and Calin Belta addressed this in Sufficient Conditions for Feasibility of Optimal Control Problems Using Control Barrier Functions by introducing a "feasibility constraint" directly into the optimization. By using High-Order CBFs, they mathematically ensure that the system never enters a state where the required safety intervention exceeds the hardware's physical capabilities.

Learning the Limits

For complex systems like the F-16 jet or multi-agent drone swarms, hand-crafting these functions is nearly impossible. Oswin So, Chuchu Fan, and others developed Policy Neural Control Barrier Functions (PNCBFs) in How to Train Your Neural Control Barrier Function: Learning Safety Filters for Complex Input-Constrained Systems. This approach reframes safety as a learning problem, training a neural network to find a safety filter that respects strict actuator limits even in high-dimensional state spaces.

The Geometry of Saturation

Why do these controllers "blow up" near the boundary? Mohammed Alyaseen and Jorge Cortés explain in Continuity and Boundedness of Minimum-Norm CBF-Safe Controllers that the mathematical geometry of the safe set can force control signals toward infinity as the system approaches the edge.

To solve this, researchers like Han Wang and Kostas Margellos propose Co-Design strategies. In Convex Co-Design of Control Barrier Function and Safe Feedback Controller Under Input Constraints, they use Sum-of-Squares (SOS) programming to simultaneously design the safety function and the controller. By treating the actuator limits as a fundamental part of the geometry from the start, they create safe regions that are practically reachable rather than just theoretically valid.

Real-World Robustness

The transition from theory to practice introduces further complications:

By integrating physical constraints—like the steering torque limits in lane-keeping systems Hao Zhou or the fuel consumption limits in gas turbines Georgy Makaryants—directly into the safety logic, this field ensures that "provable safety" survives the contact with real-world hardware.

Papers in this version

History & citationVersion 1 · 2026-10-04

Split into other directions · Permanent link

View frozen version 1

Cite this direction

Latest version (a living page):

@misc{lacuna_safety_critical_control_under_physical_i_7d384b55,
  title = {Safety-Critical Control under Physical Input Limits},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/safety-critical-control-under-physical-input-limits/rdr_7d384b554c754a7dac5dda2e7d6651b6},
  note = {Latest version of a living research direction; cite a version for a fixed text. Accessed 2026-11-20},
}

This exact version (version 1, frozen text):

@misc{lacuna_safety_critical_control_under_physical_i_7d384b55_v1,
  title = {Safety-Critical Control under Physical Input Limits},
  author = {{Lacuna}},
  howpublished = {Lacuna research direction},
  year = {2026},
  url = {https://lacuna.tiptreesystems.com/direction/safety-critical-control-under-physical-input-limits/rdr_7d384b554c754a7dac5dda2e7d6651b6/v/1},
  note = {Version 1, frozen 2026-10-04. Accessed 2026-11-20},
}